Gen3 Gt Api Runbooks Crewai - GT-Edge-AI/GT-AI-OS GitHub Wiki

Start Here

Point CrewAI OpenAI LLM configuration at GT API for agent loops; keep GT conversation creation and file handling in the orchestration layer wrapping each crew run. Recommended preset: Conversation app with uploads.

Why this matters

This runbook maps CrewAI to GT API routes operators actually publish.

Details

Compatibility: OpenAI plus GT extensions · Category: Framework and orchestration

Official documentation

• https://docs.crewai.com/concepts/llms
• https://docs.crewai.com/how-to/llm-connections
• https://docs.crewai.com/concepts/agents

Configuration fields

• OPENAI_API_BASE / OPENAI_BASE_URL: https://<tenant-host>/api/tenant
• OPENAI_API_KEY: gtak_...
• Agent LLM model: published-agent-alias

GT route mapping

GT route	Verdict	Client integration
GET /v1/models	native	Crew reads configured model string; verify allowlist with manual model list call
POST /v1/chat/completions	native	CrewAI OpenAI LLM backend posts to /v1/chat/completions
POST /v1/embeddings	not_supported	Not default CrewAI path — add custom tooling if embed scope is required
POST /v1/audio/transcriptions	not_supported	Not supported via standard CrewAI OpenAI LLM config
POST /v1/audio/speech	not_supported	Not supported via standard CrewAI OpenAI LLM config
POST /v1/images/generations	not_supported	Not supported unless custom tasks call image route
POST /v1/conversations/files	gt_extension	Custom crew tool HTTP multipart with GT headers
POST /v1/datasets/{id}/files	gt_extension	Custom crew tool targeting dataset upload route
GET /v1/files/{id}	gt_extension	Custom tool polling /v1/files/{id}

Not supported in this product

• CrewAI does not natively manage GT conversation headers or dataset multipart uploads.

Prerequisites

• Publish the alias agents should share across crew roles.
• Decide stateless vs GT conversation-per-session before enabling uploads.

Setup steps

1. Export OPENAI_API_BASE=${GT_BASE} and OPENAI_API_KEY=gtak_... in the crew runner environment.
2. Configure each agent LLM with the published alias model name.
3. Run a single-task crew completion before multi-agent autonomy.
4. If files are required, add a custom tool that calls GT upload routes with bearer auth.

GT extensions and caveats

• Store GT conversation ids in crew session state external to CrewAI memory defaults.
• Use /v1/conversations/files when uploads must follow the same GT thread as chat.

Validation checklist

• All crew roles complete with the same published alias boundary.
• GT conversation id lifecycle is logged when persistence is enabled.
• Upload tasks use GT routes, not assumed OpenAI vendor file APIs.

Plain-text export

CrewAI runbook
OpenAI plus GT extensions · Framework and orchestration
Recommended key preset: Conversation app with uploads
Evidence: documented compatibility (vendor docs cross-check)

Point CrewAI OpenAI LLM configuration at GT API for agent loops; keep GT conversation creation and file handling in the orchestration layer wrapping each crew run.

Official documentation:
- https://docs.crewai.com/concepts/llms
- https://docs.crewai.com/how-to/llm-connections
- https://docs.crewai.com/concepts/agents

Configuration fields:
- OPENAI_API_BASE / OPENAI_BASE_URL: https://<tenant-host>/api/tenant
- OPENAI_API_KEY: gtak_...
- Agent LLM model: published-agent-alias

GT route mapping:
- GET /v1/models (native): Crew reads configured model string; verify allowlist with manual model list call
- POST /v1/chat/completions (native): CrewAI OpenAI LLM backend posts to `/v1/chat/completions`
- POST /v1/embeddings (not_supported): Not default CrewAI path — add custom tooling if embed scope is required
- POST /v1/audio/transcriptions (not_supported): Not supported via standard CrewAI OpenAI LLM config
- POST /v1/audio/speech (not_supported): Not supported via standard CrewAI OpenAI LLM config
- POST /v1/images/generations (not_supported): Not supported unless custom tasks call image route
- POST /v1/conversations/files (gt_extension): Custom crew tool HTTP multipart with GT headers
- POST /v1/datasets/{id}/files (gt_extension): Custom crew tool targeting dataset upload route
- GET /v1/files/{id} (gt_extension): Custom tool polling `/v1/files/{id}`

Not supported in this product:
- CrewAI does not natively manage GT conversation headers or dataset multipart uploads.

Prerequisites:
- Publish the alias agents should share across crew roles.
- Decide stateless vs GT conversation-per-session before enabling uploads.

Setup steps:
1. Export `OPENAI_API_BASE=${GT_BASE}` and `OPENAI_API_KEY=gtak_...` in the crew runner environment.
2. Configure each agent LLM with the published alias model name.
3. Run a single-task crew completion before multi-agent autonomy.
4. If files are required, add a custom tool that calls GT upload routes with bearer auth.

GT extensions and caveats:
- Store GT conversation ids in crew session state external to CrewAI memory defaults.
- Use `/v1/conversations/files` when uploads must follow the same GT thread as chat.

Validation checklist:
- All crew roles complete with the same published alias boundary.
- GT conversation id lifecycle is logged when persistence is enabled.
- Upload tasks use GT routes, not assumed OpenAI vendor file APIs.